Title: Solvency games: How to gamble forever

                 Nevin Kapur
   Center for the Mathematics of Information
       California Institute of Technology

ABSTRACT:  An investor has a choice of several actions available to him,
each of which is represented as a probability distribution on integers (the
gain or loss that results from taking an action).  What strategy should a
risk-averse investor (one that wishes to minimize the probability of ever
going bankrupt) employ?

In this talk, I will show the existence of optimal strategies, characterize
optimal strategies, and present algorithms for determining them.  Though I
will draw upon some material about Markov Decision Processes presented last
week, this talk will be self-contained.

This is joint work with Noam Berger, Leonard Schulman, and Vijay Vazirani.